TIS-100 Bench: Dynamic Pattern Detector

Read a pattern, then detect it in a second stream. The pattern is supplied at runtime rather than built into the program.

In TIS-100 Bench, 16 LLMs each had one attempt at this TIS-100 puzzle, with 45 minutes to find a working program and keep improving it, without seeing the community leaderboard. 12 of 16 models solved it. Together they found 16 new Pareto-frontier points beyond the community leaderboard.

12 / 16models solved it
209best LLM cycles (166 community)
16new frontier points

Records

ScoreCommunity bestBest LLMModel
Cycles166209Gemini 4 Argon · +26%
Nodes55GPT-6 Astra Ties record
Instructions2331Sonnet 5.5 · +35%

Community records from the community leaderboard when we froze it. A record counts only after fresh-game replay and 100,000 passing random tests.

Pareto frontier

Pareto explorer

Lower is better ↙
Loading interactive Pareto explorer…
Community leaderboardModel frontier pointNew frontier point
Select a point with the mouse or keyboard to inspect its scores and program.

Search over time

Loading puzzle trajectories…

Every model

ModelSolvedBest cyclesBest nodesBest instructionsNew frontier points
01Gemini 4 ArgonYes20953313
02Opus 5.5Yes21563410
03Sonnet 5.5Yes2816313
04GPT-6.1 SolYes242739—
05Gemini 3.8 FlashYes245742—
06Fable 5.1Yes289640—
07GPT-6 AstraYes297538—
08DeepSeek V4.1 FlashYes321749—
09Haiku 5.5Yes327739—
10Grok 4.7Yes409856—
11Muse Wonder 1.0Yes626967—
12Kimi K3.1 ProYes1,106773—
13GLM-5.3No————
14MiMo-V2.6-ProNo————
15Muse Spark 1.3No————
16GPT-6 LunaNo————

Best scores can come from different programs. New frontier points are not matched or beaten on all three scores by any community program.

Best model programs

These programs are in TIS-100 save format and released under the MIT License. All of this puzzle’s model programs are on GitHub and Hugging Face.

Gemini 4 Argon Fewest cycles★ New frontier point

209 cycles · 7 nodes · 44 instructionsNEXUS.16.212.8.209-7-44.txt on GitHub

Gemini 4 Argon program running on NEXUS.16.212.8 in TIS-100
Node 0
MOV UP, ACC
MOV UP, RIGHT
MOV UP, RIGHT
L: MOV ACC, RIGHT
JMP L
Node 1
MOV LEFT, DOWN
MOV LEFT, DOWN
JMP C
Z: MOV -1, DOWN
C: MOV RIGHT, ACC
SUB LEFT
JNZ Z
M: MOV 1, DOWN
MOV RIGHT, ACC
SUB LEFT
JNZ Z
JMP M
Node 2
MOV UP, ACC
MOV ACC, DOWN
MOV ACC, LEFT
Node 5
MOV UP, ACC
MOV UP, DOWN
Z: MOV 0, RIGHT
L: JRO UP
MOV ACC, RIGHT
JMP L
Node 6
Z: MOV UP, ACC
MOV 0, DOWN
SUB LEFT
JNZ Z
M: MOV UP, ACC
MOV ACC, DOWN
SUB LEFT
JNZ Z
JMP M
Node 9
MOV UP, ACC
L: MOV ACC, RIGHT
JMP L
Node 10
MOV UP, ACC
SUB LEFT
JNZ Z0
MOV 1, DOWN
JMP E
Z0: MOV 0, DOWN
E:

Gemini 4 Argon Fewest nodes

288 cycles · 5 nodes · 41 instructionsNEXUS.16.212.8.288-5-41.txt on GitHub

Gemini 4 Argon program running on NEXUS.16.212.8 in TIS-100
Node 0
MOV UP, RIGHT
MOV UP, ACC
MOV UP, RIGHT
L: MOV ACC, RIGHT
JMP L
Node 1
MOV LEFT, ACC
MOV LEFT, RIGHT
SAV
Z: SWP
SAV
SUB RIGHT
MOV 0, RIGHT
JNZ Z
M: SWP
SAV
SUB RIGHT
MOV LEFT, RIGHT
JNZ Z
JMP M
Node 2
MOV LEFT, DOWN
Z: MOV UP, ACC
MOV ACC, LEFT
MOV 0, DOWN
SUB LEFT
JNZ Z
M: MOV UP, ACC
MOV ACC, LEFT
MOV ACC, DOWN
SUB LEFT
JNZ Z
JMP M
Node 6
MOV UP, ACC
L: MOV UP, DOWN
MOV ACC, DOWN
JMP L
Node 10
MOV UP, ACC
SUB UP
JNZ Z
MOV 1, DOWN
JMP E
Z: MOV 0, DOWN
E:

Sonnet 5.5 Fewest instructions★ New frontier point

370 cycles · 6 nodes · 31 instructionsNEXUS.16.212.8.370-6-31.txt on GitHub

Sonnet 5.5 program running on NEXUS.16.212.8 in TIS-100
Node 0
MOV UP, RIGHT
Node 1
MOV LEFT, ACC
SWP
MOV LEFT, ACC
MOV LEFT, DOWN
L: SWP
MOV ACC, RIGHT
JMP L
Node 2
D: MOV UP, ACC
MOV 0, DOWN
B: SWP
SUB LEFT
JNZ K
SWP
SAV
SUB LEFT
JNZ D
A: MOV UP, ACC
MOV ACC, DOWN
JMP B
K: MOV LEFT, NIL
Node 5
MOV UP, ACC
L: MOV ACC, RIGHT
JMP L
Node 6
S: MOV UP, ACC
SUB LEFT
JNZ Z
MOV 1, DOWN
JMP S
Z: MOV 0, DOWN
Node 10
MOV UP, DOWN